Update README.md

Merge pull request intel#38 from 01org/develop

Develop

Merge pull request intel#36 from 01org/develop

Develop

Merge pull request intel#35 from 01org/develop

Merge from the development branch. Major changes include sub-blocks and default args.

Merge branch 'master' of https://github.com/01org/yask

Conflicts:
	docs/YASK-intro.pdf

Sub domains (intel#18)

* Use pointers to grids to improve code-gen in some stencils.

Grids as member vars instead of pointers prevented the compiler from optimizing addr calculations, causing >3x code bloat in AWP stencil and breaking vectorization in FSG.

Also:
Fixed uninitialized vars in grids.
Fixed some g++ compilation warnings.
Ensured cache-line-boundary alignment in allocation.

* Bugfix typo (prevented compilation). (intel#10) (intel#12)

* Sync with master (intel#17)

* Sub domains (intel#13)

* Use pointers to grids to improve code-gen in some stencils.

Grids as member vars instead of pointers prevented the compiler from optimizing addr calculations, causing >3x code bloat in AWP stencil and breaking vectorization in FSG.

Also:
Fixed uninitialized vars in grids.
Fixed some g++ compilation warnings.
Ensured cache-line-boundary alignment in allocation.

* Bugfix typo (prevented compilation). (intel#10) (intel#12)

* Sub domains (intel#14)

* Use pointers to grids to improve code-gen in some stencils.

Grids as member vars instead of pointers prevented the compiler from optimizing addr calculations, causing >3x code bloat in AWP stencil and breaking vectorization in FSG.

Also:
Fixed uninitialized vars in grids.
Fixed some g++ compilation warnings.
Ensured cache-line-boundary alignment in allocation.

* Bugfix typo (prevented compilation). (intel#10) (intel#12)

* Sub domains (intel#15)

* Use pointers to grids to improve code-gen in some stencils.

Grids as member vars instead of pointers prevented the compiler from optimizing addr calculations, causing >3x code bloat in AWP stencil and breaking vectorization in FSG.

Also:
Fixed uninitialized vars in grids.
Fixed some g++ compilation warnings.
Ensured cache-line-boundary alignment in allocation.

* Bugfix typo (prevented compilation). (intel#10) (intel#12)

* Update intro doc.

BDW and multi-node results TBD.

Update perf and MPI docs.

Change build defaults

Add support for Intel OpenMP hot teams.

Hot teams is now the default.
Crew still supported, but disabled by default.

Change some default settings.

Make prefetch to L1 on by default.
Turn commutative recomb off by default.
Change default L2 prefetch to 2.
Add 'vector always' to simd loops.
Add ./lib to lib path.